DeepSeek-V3.2 vs DeepSeek-V4-Flash
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
DeepSeek-V4-Flash is the stronger all-rounder.DeepSeek-V3.2 is cheaper.
Scores updated · 38 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software07DeepSeek-V4-Flash7 of 8 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own04DeepSeek-V4-Flash4 of 4 tests
- ReasoningHard problems that need careful thinking04DeepSeek-V4-Flash4 of 4 tests
- MathCompetition and research-level math03DeepSeek-V4-Flash3 of 3 tests
- FactsGetting facts right instead of making them up12DeepSeek-V4-Flash2 of 3 tests
- Long documentsFinding answers in very long texts01DeepSeek-V4-Flash1 of 1 test
- Following instructionsDoing exactly what it is asked01DeepSeek-V4-Flash1 of 1 test
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
DeepSeek-V3.2 costs 60% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where DeepSeek-V4-Flash pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+57.4points ahead
- Real work tasks from 44 professionsGDPVal+37.1points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+37points ahead
Where DeepSeek-V3.2 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+9points ahead
Every test, side by side
All 38 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Flash
- SWE-bench ProDeepSeek-V4-Flash by 3715.652.6+37
- Terminal-Bench 2.1DeepSeek-V4-Flash by 31.946.878.7+31.9
- SciCodeDeepSeek-V4-Flash by 11.438.950.3+11.4
- LiveCodeBench v6DeepSeek-V4-Flash by 7.383.390.6+7.3
- LMArena · WebDevDeepSeek-V4-Flash by 68 rating points13621430+68 rating
- SWE-bench VerifiedDeepSeek-V4-Flash by 5.973.179+5.9
- SWE-bench MultilingualDeepSeek-V4-Flash by 3.170.273.3+3.1
- Terminal-Bench Hardtie35.635.6tie
AgentsDeepSeek-V4-Flash
- GDPValDeepSeek-V4-Flash by 37.19.846.9+37.1
- BrowseCompDeepSeek-V4-Flash by 21.851.473.2+21.8
- τ-Bench V3 · BankingDeepSeek-V4-Flash by 20.618.839.4+20.6
- MCP AtlasDeepSeek-V4-Flash by 6.862.269+6.8
ReasoningDeepSeek-V4-Flash
- ARC-AGI-2DeepSeek-V4-Flash by 57.4461.4+57.4
- Humanity's Last ExamDeepSeek-V4-Flash by 1424.638.6+14
- CritPtDeepSeek-V4-Flash by 13.72.916.6+13.7
- GPQA DiamondDeepSeek-V4-Flash by 6.88490.8+6.8
FactsDeepSeek-V4-Flash
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2 by 917.38.3+9
- AA-Omniscience · AccuracyDeepSeek-V4-Flash by 7.43340.4+7.4
- SimpleQA VerifiedDeepSeek-V4-Flash by 6.627.534.1+6.6
MathDeepSeek-V4-Flash
- HMMT Feb. 2026DeepSeek-V4-Flash by 10.784.194.8+10.7
- IMOAnswerBenchDeepSeek-V4-Flash by 10.178.388.4+10.1
- AIME 2026DeepSeek-V4-Flash by 1.694.295.8+1.6
Long documentsDeepSeek-V4-Flash
- AA-LCRDeepSeek-V4-Flash by 6.473.379.7+6.4
Following instructionsDeepSeek-V4-Flash
- IFBenchDeepSeek-V4-Flash by 18.560.779.2+18.5
Other results14 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- CyberGymDeepSeek-V4-Flash by 59.417.376.7+59.4
- ToolathlonDeepSeek-V4-Flash by 35.135.270.3+35.1
- ARC-AGI-1DeepSeek-V4-Flash by 325789+32
- Artificial Analysis Coding IndexDeepSeek-V4-Flash by 24.944.269.1+24.9
- AA Agentic IndexDeepSeek-V4-Flash by 23.418.341.7+23.4
- AA IntelligenceDeepSeek-V4-Flash by 12.821.534.3+12.8
- Terminal-Bench 2.0DeepSeek-V4-Flash by 10.546.456.9+10.5
- GPQA (unspecified)DeepSeek-V4-Flash by 8.679.988.5+8.6
- LiveCodeBenchDeepSeek-V4-Flash by 8.383.391.6+8.3
- AA-OmniscienceDeepSeek-V4-Flash by 8.2-22.5-14.3+8.2
- BrowseComp (context management)DeepSeek-V4-Flash by 5.667.673.2+5.6
- τ²-Bench Telecom (AA run)DeepSeek-V4-Flash by 4.490.695+4.4
- HLE (with tools)DeepSeek-V4-Flash by 4.340.845.1+4.3
- MMLU-ProDeepSeek-V4-Flash by 1.28586.2+1.2
Questions people ask
Which is better, DeepSeek-V3.2 or DeepSeek-V4-Flash?
DeepSeek-V4-Flash wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. DeepSeek-V3.2 wins none, but costs 60% less.
Which is better for coding?
DeepSeek-V4-Flash. It wins 7 of the 8 coding tests both models report; DeepSeek-V3.2 wins none, and 1 is a tie.
Which is cheaper?
DeepSeek-V3.2 costs $0.28 per million input tokens and $0.42 per million output tokens; DeepSeek-V4-Flash costs $0.44 and $1.32. That makes DeepSeek-V3.2 about 60% cheaper for the same work.
How do you compare the two?
We use the 38 benchmark tests both models have published scores on. The verdict counts the 24 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 14 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.